KR102290372B1 - Opportunity window hints for background operations in ssd - Google Patents

Abstract

The present invention includes the steps of receiving a plurality of host commands from a host to access a storage medium of a solid-state drive (SSD), monitoring the raw speed to execute the plurality of host commands, calculating the average velocity by taking the average, comparing the average velocity to a threshold, detecting that the average velocity indicative of the opening of the opportunity window falls below the threshold, providing hints about the opportunity window; and determining whether to perform pending or impending background operations during the opportunity window.

Description

OPPORTUNITY WINDOW HINTS FOR BACKGROUND OPERATIONS IN SSD

This disclosure relates generally to solid-state drives (SSDs), and more particularly, to a system and method for providing opportunity window hints to perform a background operation in an SSD. will be.

SSDs store host/user data in a storage medium such as a flash medium. Flash media have specific read/write and endurance characteristics that must be carefully managed. The SSD controller manages the flash media of the SSD to provide reliable, high-performance, and cost-effective data storage. The SSD controller needs to perform various internal background operations in order to extend the life cycle and provide reliable operation of the SSD. Examples of such internal background operations are recycling, garbage collection, data structure updates for the Flash Translation Layer (FTL), various statistics and log updates, and cache offloads.

These background operations can take up a significant portion of the flash channel bandwidth and resources of the SSD controller. Some background operations, such as garbage collection and recycling, are very complex, potentially causing severe interruption of host read/write operations. The bandwidth and latency observed by the host can fluctuate significantly due to internal background operations of the SSD.

These background operations can take up a significant portion of the flash channel bandwidth and resources of the SSD controller. Some background operations, such as garbage collection and recycling, are very complex, potentially causing severe interruptions to host read/write operations. The bandwidth and latency observed by the host can fluctuate significantly due to internal background operations of the SSD.

According to an embodiment, a method includes: receiving a plurality of host commands from a host to access a storage medium of a solid-state drive (SSD); monitoring the raw speed to execute a plurality of host commands; calculating an average velocity by taking the average of the raw velocity over a unit of time; comparing the average speed to a threshold; detecting that the average velocity falls below the threshold, indicating an opening of a window of opportunity; providing hints for an opportunity window; and determining whether to perform pending or impending background operations during the opportunity window.

According to another embodiment, a solid-state drive (SSD) includes: a flash medium; and a controller including a host interface layer, a flash translation layer, and a flash interface layer. The host interface layer is configured to: receive a plurality of host commands from the host to access the storage medium of the SSD; monitor raw speed to execute a plurality of host commands; calculate the average velocity by taking the average of the raw velocity over time units; compare the average speed to a threshold; detecting that the average velocity falls below the threshold, indicating the opening of a window of opportunity; and provide hints for the opportunity window to the flash transformation layer. The flash transformation layer is configured to determine whether to perform pending or impending background operations during the opportunity window. The flash interface layer is configured to perform data transfer to or from the flash medium according to a plurality of host commands.

BRIEF DESCRIPTION OF THE DRAWINGS The above and other preferred features, including various novel details of implementation and combination of events, will be more particularly described with reference to the appended drawings and pointed out in the claims. Certain systems and methods described herein are presented by way of illustration, but not limitation. As will be understood by one of ordinary skill in the art, the principles and features described herein may be applied in various and numerous embodiments without departing from the scope of the present disclosure.

The present disclosure provides a system and method for detecting and performing optimal time windows for performing background operations of a solid-state drive (SSD). The SSD controller provides data for recycling, garbage collection, and Flash Translation Layer (FTL) in optimal time windows to minimize interruption and avoid potential conflicts with host-initiated I/O operations. It can perform internal background operations of the SSD such as structure updates, various statistics and log updates, and cache offloads.

1 shows a block diagram of an exemplary solid-state drive (SSD), according to an embodiment;
2 is a flow diagram of an exemplary process for calculating an average host read speed, according to one embodiment;
3 shows an example of an opportunity window;
4 shows an example use case of a database application to show windows of opportunity for performing background operations; and
5 is a flow diagram of an example process for performing background operations based on an average host read speed, according to one embodiment.

Each of the features and teachings disclosed herein may be used in conjunction with or separately from other features and teachings to provide for performing background operations in an SSD. Representative examples in which these additional features and teachings are used separately or in combination are described in detail with reference to the accompanying drawings. This detailed description is only intended to teach one skilled in the art for practicing aspects of the present teachings, and does not limit the scope of the claims. Therefore, combinations of features recited in the Detailed Description may not need to practice the teachings in a broad sense, but are instead taught merely for the purpose of illustrating the present teachings, particularly representative embodiments.

In the following description, for purposes of explanation only, specific designations are set forth in order to provide a thorough understanding of the present invention. However, it will be apparent to those skilled in the art that these specific details are not required to practice the spirit of the present invention.

Some portions of the detailed description are presented in terms of algorithms and symbolic representations of operations on data bits in computer memory. These algorithmic descriptions and representations are used by those skilled in the data processing arts to effectively convey the substance of their work to those skilled in the art. Here the algorithm can generally be a consistent sequence of steps leading to a desired result. Steps are those requiring physical manipulation of physical quantities. Generally, though not necessarily, these quantities take the form of electrical or magnetic signals capable of being stored, transferred, combined, compared, and otherwise manipulated. It has proven convenient at times, principally for reasons of common usage, to refer to these signals as bits, values, elements, symbols, features, terms, numbers, and the like.

It should be borne in mind, however, that all of these and similar terms are to be associated with the appropriate physical quantities and are merely convenient labels applied to these quantities. Specifically, as will be apparent from the description below, in the description, discussions using terms such as “processing”, “computing”, “compute”, “determining”, “indicating” and the like refer to computer systems or computer registers and memories. Activities and processes of similar electronic computing devices that manipulate and transform data represented as physical (electrical) quantities within computer system memories or registers or other similar data represented as physical quantities within computer system memories or registers or other information storage, transmission or display devices. is understood to represent

In addition, various features of the representative examples and dependent claims may be combined in non-explicit and non-listed ways to provide useful additional embodiments of the teachings of the present invention. It is also noted that all implied values of ranges or groups of entities disclose all possible intermediate values or intermediate entities for the purpose of limiting the invention as well as the original disclosure purposes. In addition, it should be noted that the shapes of the components shown in the explicit reference and drawings are designed to be understood in the manner implemented in the present specification, but are not limited to the dimensions and shapes shown in the embodiments.

The present disclosure provides a system and method for detecting and providing optimal time windows for performing background operations of a solid-state drive (SSD). The SSD controller provides data for recycling, garbage collection, and Flash Translation Layer (FTL) in optimal time windows to minimize interruption and avoid potential conflicts with host-initiated I/O operations. It can perform internal background operations of the SSD such as structure updates, various statistics and log updates, and cache offloads.

According to one embodiment, the host interface layer of the SSD monitors the speed of host I/O operations and provides a hint to the flash translation layer of the SSD indicating that optimal time windows are available. Optimal time windows are also referred to as opportunity windows. The FTL may determine whether to execute some or all of the pending or impending background operations in optimal time windows, reducing or avoiding impact on host-initiated I/O operations. The host interface layer does not propose any specific flash conversion algorithm, only based on the existing FTL algorithm, may provide hints on when to perform background operations to the FTL.

1 shows a block diagram of an exemplary solid-state drive (SSD), according to an embodiment. The SSD 120 includes a flash medium 150 and an SSD controller 121 . According to an embodiment, the SSD 120 is a Non-Volatile Memory express (NVMe) memory including a non-volatile storage medium attached via a Peripheral Component Interconnect express (PCIe) bus. In another embodiment, the SSD 120 is a memory compatible with the NVMe over Fabrics (NVMe-oF) standard. In this case, the SSD 120 may be accessed via underlying fabrics such as Ethernet, Fiber Channel, and InfiniBand.

The SSD controller 121 includes a host interface layer 122 , a flash translation layer (FTL) 123 , and a flash translation layer 124 . The host interface layer 122 provides a host interface between the host 110 and the SSD controller 121 . The FTL 123 converts Logical Block Addresses (LBAs) to Physical Block Addresses (PBAs) using a mapping table, and the requested host I/ To perform O commands, the physical block of the flash match 150 is accessed as indicated in the mapping table. Examples of host I/O commands include, but are not limited to, read, write, read/modify/write (RMW), and delete commands. Flash interface layer 124 performs the actual data transfer to or from flash medium 150 according to host I/O commands and FTL mappings.

According to an embodiment, the host interface layer 122 may track the speed of host read operations. For example, the host interface layer 122 calculates the host read rate by averaging the host read operations over a predetermined period of time (eg, 1 microsecond, 1 millisecond, 1 second). Host read speed is also referred to herein as host read bandwidth. The host interface layer 122 periodically checks the host read speed against a threshold (ie, the host read speed threshold). The host read speed threshold may be programmed through a command received from the host 110 . When the host read rate falls below the host read rate threshold, the host interface layer 122 considers that the background operation opportunity window is open. If the host read rate exceeds the host read rate threshold, the host interface layer 122 considers that the background operation opportunity window is closed.

The FTL 123 may initiate and perform house-keeping background operations during the opportunity window. For example, the FTL 123 may update the FTL mapping and perform a background operation for storing the updated FTL mapping table in the flash medium 150 . As host commands are executed, the FTL mapping table is continuously changed as new write data is written to new physical locations of the flash medium 150 . The FTL mapping table may change due to background operations such as offloading the write cache, recycling and garbage collection of flash media, and LBA trim commands from the host. The FTL 123 needs to periodically store changes in the FTL mapping table to the flash medium 150 in a background operation. In addition to FTL mapping, FTL 123 updates and periodically stores other FTL statistics tables, such as program/erase (P/E) counts per block and various logs, to flash medium 150 .

Background Operations During an opportunity window, the FTL 123 may perform all or some of the background operations as long as the opportunity window is open. When the opportunity window is closed, the FTL 123 may suspend or reduce active background operations to minimize the impact on subsequent host I/O performance.

According to an embodiment, the FTL 123 may select to perform specific background operations among pending or imminent background operations during the opportunity window. Depending on the nature of the application, threshold, and workload, windows of opportunity may not be available or difficult to obtain periodically. In this case, the FTL 123 may perform certain pending or imminent background operations regardless of opportunity window hints received from the host interface layer 122 . In an embodiment, the SSD controller 121 may determine that the current threshold is set too high to have a sufficient probability or duration of opportunity windows. In this case, the SSD controller 121 may use a threshold value from a predetermined setting of threshold values set by the host. Certain background operations may be delayed, but not indefinitely, and must be performed within a certain period of time to achieve the required reliability and durability of SSD 120 while not impacting overall host I/O performance. In this case, the window of opportunity may only represent a window of time that represents a potential opportunity to perform background operations when less host I/O operations are being performed, but does not require the operations to occur during this window. Therefore, background operations selectively performed within the opportunity windows can reduce or avoid conflict or collision with host operations and minimize the impact on host I/O performance.

Although the host read speed is mainly used for opportunity window detection, the SSD controller 121 may track different speeds of host operations to detect the opportunity window. For example, the SSD controller 121 may track the speed of host write operations and detect an opportunity window based on the host write speed in a manner similar to the host read speed. The threshold used to detect the window of opportunity based on the host write speed may be different from the threshold used to detect the window of opportunity based on the host read speed. In some embodiments, the SSD controller 121 may track both the host read speed and the host write speed to detect an opportunity window to balance both the host read speed and the host write speed. In this case, a weighting factor may be applied when calculating the summed host read/write speed for explaining the contribution of each of the host read and write operations to the overall host I/O performance. In some embodiments, the host read rate may be used for a host read intensive application, and the host write rate may be used for a host write intensive application.

According to one embodiment, the host interface layer 122 may track multiple rates of host operations. For example, the host interface layer 122 may track both the rate of host read operations and the rate of host write operations. The host interface layer 122 may periodically check multiple rates of host operations for a set of rate thresholds (eg, a host read rate threshold and a host write rate threshold). The speed thresholds may be programmed through a command received from the host 110 . When each of the speeds falls below the corresponding speed threshold, the host interface layer 122 considers an opportunity window open. According to an embodiment, multiple speeds of host operations may be checked independently of each other. This is because the resources required to perform some background operations may be different from the resources required to perform other background operations. In this case, the opportunity window for specific background operations may be opened while the opportunity window for other background operations is closed. According to another embodiment, multiple velocities are simultaneously checked against corresponding velocity thresholds to determine whether the opportunity window is open or closed.

The host interface layer 122 performs host command fetching, initiates execution of host commands, transmits data related to the host commands, and finally completes to indicate whether the host commands were successfully executed. Post items to host 110 . In addition to these general tasks, the host interface layer 122 may monitor the host read/write speed as each host command is completed. The host interface layer 122 may accumulate a size (eg, bytes, doublewords, or logical blocks) of data transmitted from or to the host 110 during a unit of time. can Some examples of time unit resolution are 1 microsecond, 1 millisecond, and 1 second. The time unit for accumulating data transmission may be programmed by the host 110 through a host command. For every unit of time, the host interface layer 122 keeps track of the raw host read/write speed based on the size of the accumulated data. When calculating the raw host read/write speed, the host interface layer 122 may apply a smoothing function (eg, a low-pass filter). Based on the raw host read/write speed, the host interface layer 122 calculates an average host read/write speed by applying a programmable weight to the raw host read/write speed. The raw host read/write speed and the calculated average host read/write speed may be stored for detecting opportunity windows and/or for future reference. The raw host read/write speed can fluctuate rapidly, and the host interface layer 122 can average the raw host read/write speed to smooth out the sharpness and calculate the average host read/write speed. there is. The weight of the average function may be programmed and fine-tuned by system software and/or host driver based on the application running on the host 110 .

The following is pseudo code for calculating a raw host read/write speed and an average host read/write speed, according to an embodiment.

For every host command { // compute raw rate

Fetch host command;

Initiate host command execution;

Perform any data transfers needed;

Post appropriate completion entry to the host;

If host command is read/write operation {

accumulated_read/write_size = accumulated_read/write_size + currentread/write command size;

}

}

For every time trigger { // compute average rate

avg _read/write_rate (new) = WEIGHT * accumulated_read/write_size + (1 - WEGHT) * avg_read/write_rate (old);

accumulated_read/write_size = 0; // reset it

}

For each host read/write command, the host interface layer 122 calculates an accumulated data size corresponding to the executed host commands. The accumulated read/write size values are stored in an accumulator.

For every time trigger (eg, 1 microsecond), the host interface layer 122 calculates the average host read/write speed based on the average weight parameter (WEIGTH). The average weight parameter (WEIGTH) is programmatically set between 0 and 1. The average weight parameter (WEIGTH) is used to determine the degree of flattening to be performed for the raw host read rate. For example, a weight parameter (WEIGTH) of “1” indicates no flattening, while a weight parameter of WEIGTH of “0” indicates that the average is clamped to zero. After calculating the average host read/write speed, the host interface 122 resets the accumulated read/write size of the accumulator to calculate a new average host read/write speed at a subsequent time unit. The host interface layer 122 may repeat the process and continuously calculate, monitor, and store values of the host read/write speed. The host interface layer 122 periodically compares the values of the host read/write speed with one or more threshold values, and provides hints related to windows of opportunity for performing background operations to the FTL 123 .

2 is a flow diagram of an exemplary process for calculating an average host read speed, according to one embodiment. Note that a similar flow diagram is applicable for host write rates or combined host read/write rates.

The host issues I/O commands to the SSD (201). The host interface layer of the SSD controller of the SSD performs command fetching ( 202 ) and starts executing the received host I/O commands ( 203 ). The flash translation layer (FTL) and flash interface layer of the SSD controller perform data transfers 204 to execute each of the host I/O commands, and the host interface layer finally posts a completion input to the host. For each executed host command, the host interface layer checks whether the executed host command is a host read command (206), and updates the host read data size (207). For other host commands, steps 206 and 207 for calculating the average host read rate are omitted. For each executed host command, the host interface layer checks whether the executed host command is a host read command (206), and updates the host read data size (207). For other host commands, steps 206 and 207 for calculating the average host read rate are omitted.

The host interface layer also executes a timer (211). For every time trigger (eg, 1 microsecond), the host interface layer computes and updates the host read rate 212 , and computes a new host read/write rate corresponding to the subsequent unit of time. The read size is reset (213).

Write cache offloading is another example of internal background operations performed by the FTL 123 . The write cache may reside in on-chip buffers, external memory, or even faster flash-type media. The FTL 123 periodically sends a write cache to the primary flash medium 150 . When the FTL 123 performs a write cache offload operation during a peak time window in which I/O operations are intensively performed, the performance of host I/O operations may be adversely affected.

The FTL 123 may also need to periodically perform recycling and garbage collection of flash blocks/pages. Flash blocks of flash medium 150 need to be erased and added to the pool of free blocks. Some flash blocks selected for garbage collection may contain some valid data that must be moved to another flash block(s) before the blocks are erased. Since these recycling and garbage collection operations may occupy a large amount of flash bandwidth, host I/O performance may be adversely affected when host I/O operations conflict or collide with recycling and garbage collection operations.

According to an embodiment, the host interface layer 122 may monitor the average host read/write speed and compare the average host read/write speed with a programmable threshold value. When the average read/write speed falls below a set threshold, the host interface layer 122 considers the background operation opportunity window to be open. The host interface layer 122 provides an opportunity window to the FTL 123 as a hint for performing background operations. Using the opportunity window, the FTL 123 determines the actual execution of background operations. Background operations scheduled during the opportunity window will have minimal impact on host I/O performance.

According to an embodiment, the FTL 123 may ignore the opportunity window hints received from the host interface layer 122 and may not perform specific background operations during the opportunity window for various reasons. For example, some impending background operations may need to be performed after a certain event or during a certain time window, regardless of hints to the opportunity window. When such an opportunity window hint is received, the FTL 123 may verify that all or some pending or impending background operations can be performed. When the host read/write speed exceeds the threshold, the host interface layer 122 indicates to the FTL 123 that the window of opportunity is closed. After the opportunity window is closed, the FTL 123 may stop or reduce internal background operations started in the opportunity window.

3 shows an example of an opportunity window. The raw host read rate fluctuates with a high frequency, and the host interface layer averages the fluctuating raw host read rate to calculate the average host read rate. In this example, opportunity windows open between 24 and 49 microseconds and between 64 and 83 microseconds when the average host read speed is below a threshold of 32 MBps. Referring to FIG. 1 , the threshold used to detect an opportunity window is programmable and may be set by the host 110 . Host 110 may apply different threshold values based on the application and characteristics of flash medium 150 and host I/O operations. According to an embodiment, the host 110 may program various parameters to calculate the host read/write speed. Examples of host programmable parameters include, but are not limited to, opportunity window thresholds and read/write average function weights. The host 110 may program these parameters by using values stored in a designated space in registers of the SSD or by transmitting NVMe-based Set Feature/Get Feature commands or vendor-specific commands to the SSD. In some embodiments, the SSD may employ various learning algorithms to adjust parameters for calculating host read/write speed and fine-tune speed thresholds.

4 shows an example use case of a database application to show windows of opportunity for performing background operations. The workload used in this example was tailored to the RocksDB database application for the Twitter use case. In this example, the I/O size is 4 KB, the workload is reads and random writes, the database block cache is 20 GB, and 16 I/O threads are used.

The record of the host read rate shown in Figure 4 indicates that the window of opportunity is periodically available. By using opportunity windows as hints to perform background operations, the SSD controller can reduce conflicts between host I/O operations and internal background operations.

Below is the pseudocode for the FTL to use the opportunity window hints provided by the host interface layer.

For every time trigger { // check average host read rate

If (avg_read_rate < THRESHOLD) { // window opens

"Background Ops Window" = OPEN;

Perform any pending or imminent background operations;

}

Else if (avg_read_rate > THRESHOLD) { // window closes

"Background Ops Window" = CLOSE;

Halt or scale-down active background operations;

}

}

5 is a flow diagram of an example process for performing background operations based on an average host read speed, according to one embodiment. Note that a similar flow diagram is applicable for calculating host write rates or combined host read/write rates.

The host interface layer continuously tracks the host read speed and compares the host read speed with a programmable threshold (501). If the host read speed is less than the threshold value ( 502 ), the host interface layer determines that an opportunity window for performing background operations is open. In this case, the FTL performs (503) pending or impending background operations during the opportunity window. When the host read speed is greater than the threshold value ( 502 ), the host interface layer determines that the opportunity window for performing background operations is closed. In this case, the FTL stops or reduces active background operations ( 504 ) to free up flash channel bandwidth for performing host I/O operations.

Erasure Code (EC) is a method of data protection in which data is decomposed into pieces, expanded and encrypted into redundant data pieces, and stored in different locations or on different sets of storage media. Erasure coding enables reconstruction of corrupted data at some point in the disk storage process using redundant data stored in different locations or on different storage media. Erasing codes have advantages over traditional RAID in their ability to reduce the time and overhead required for data reconstruction. The problem with erasure coding is that it can translate to increased I/O latency, making it more CPU-intensive.

According to one embodiment, the present system and method for providing windows of opportunity may be extended to erasure coding of a group of SSDs. At any given point in time, the N SSDs (N depending on the erasure coding scheme) between the group of SSDs may be offline, and the data may still be restored by the erase code. When a window of opportunity is available on a particular SSD, the SSD can check (by its own calculation) whether there are N or fewer SSDs that are not operating within the erasure coding group or are hampered by increased latency. If there are fewer than N SSDs, the SSD can safely initiate background operations. Even if the opportunity window prediction is not accurate, the erasure coding group can still control operation without delay, because data can be constructed from the remaining drives that are not affected using erasure coding.

According to an embodiment, a method includes: receiving a plurality of host commands from a host to access a storage medium of a solid-state drive (SSD); monitoring the raw speed to execute a plurality of host commands; calculating an average velocity by taking the average of the raw velocity over a unit of time; comparing the average speed to a threshold; detecting that the average velocity falls below a threshold, indicating the opening of a window of opportunity; providing hints for an opportunity window; and determining whether to perform pending or impending background operations during the opportunity window.

The method includes: initiating a host command of a plurality of host commands; posting a completion input for each of the plurality of host commands; and updating the average speed after executing each host command.

The raw rate may be a raw host read rate, and the average rate may be an average host read rate.

The raw speed may be a raw host write speed, and the average speed may be an average host write speed.

The method includes: calculating a plurality of average velocities by averaging the plurality of raw velocities over a unit of time; comparing the plurality of average speeds to a set of threshold values; detecting that a first average velocity falls below a first threshold, indicating the opening of a first window of opportunity; detecting that a second average velocity indicative of the opening of a second window of opportunity falls below a second threshold; providing hints for first and second windows of opportunity; determining whether to perform a first subset of pending or impending background operations during a first window of opportunity; and determining whether to perform a second subset of pending or impending background operations during the second window of opportunity.

The SSD's host interface layer may provide hints indicating that a window of opportunity is available to the SSD's Flash Translation Layer (FTL).

Background operations include updating the FTL mapping table to flash media, storing per-block Program/Erase (P/E) counts and various logs to flash media, offloading write cache, and flash blocks/pages. They may include one or more of recycling, and garbage collection.

The average speed can be calculated as a weighting parameter.

The threshold value may be determined based on one or more of an application running on the host, a type of SSD, and characteristics of the plurality of host commands.

The method includes: detecting that the average velocity exceeds a threshold value, indicating the closing of the opportunity window; and stopping or reducing active background operations.

The method may further include: programming one or more parameters to calculate an average speed using instructions received from the host.

According to another embodiment, a solid-state drive (SSD) includes: a flash medium; and a controller including a host interface layer, a flash translation layer, and a flash interface layer. The host interface layer is configured to: receive a plurality of host commands from the host to access the storage medium of the SSD; monitor raw speed to execute a plurality of host commands; calculate the average velocity by taking the average of the raw velocity over time units; compare the average speed to a threshold; detecting that the average velocity indicative of the opening of the window of opportunity falls below a threshold; and provide hints for the opportunity window to the flash transformation layer. The flash transformation layer is configured to determine whether to perform pending or impending background operations during the opportunity window. The flash interface layer is configured to perform data transfer to or from the flash medium according to a plurality of host commands.

The host interface layer is configured to: initiate a host command from among the plurality of host commands; post a completion input for each of the plurality of host commands; and updating the average speed after executing each host command.

The raw rate may be a raw host read rate, and the average rate may be an average host read rate.

The raw speed may be a raw host write speed, and the average speed may be an average host write speed.

The host interface layer: calculates the plurality of average velocities by averaging the plurality of raw velocities over a unit of time; compare the plurality of average velocities to a set of threshold values; detecting that a first average velocity, indicating opening of a first window of opportunity, falls below a first threshold value; detecting that a second average speed, indicating opening of a second window of opportunity, falls below a second threshold; provide hints for first and second windows of opportunity; determine whether to perform a first subset of pending or impending background operations during a first window of opportunity; and determine whether to perform a second subset of pending or impending background operations during the second window of opportunity.

Background operations update the FTL mapping table to flash media, store per-block Program/Erase (P/E) counts and various logs to flash media, offload write cache, and store flash blocks/pages. one or more of recycling, and garbage collection.

The average speed can be calculated as a weighting parameter.

The threshold value may be determined based on one or more of an application running on the host, a type of SSD, and characteristics of the plurality of host commands.

The host interface layer may be configured to detect that the average speed exceeds a threshold value, indicating the closing of the opportunity window. The flash transform layer may be configured to suspend or reduce active background operations.

One or more parameters may be programmed to calculate the average speed using commands received from the host.

The SSD may be compatible with the Non-Volatile Memory express (NVMe) standard of the NVMe-oF (Non-Volatile Memory express over Fabrics) standard.

The above embodiments have been described herein to illustrate various embodiments of a system and method implementation for providing opportunity window hints for performing background operations in an SSD. Various modifications and departures from the disclosed embodiments will occur to those skilled in the art. Subject matter intended to be within the scope of the invention is set forth in the following claims.

Claims (10)

Receiving a plurality of host commands from a host to access a storage medium of a solid-state drive (SSD);
monitoring a raw speed to execute the plurality of host commands based on the amount of data transferred between the host and the SSD;
calculating an average velocity by averaging the raw velocity over a unit of time using host-programmable weights;
comparing the average speed to a threshold value;
detecting that the average velocity falls below the threshold, indicating an opening of a window of opportunity;
providing hints for the opportunity window; and
determining whether to perform pending or impending background operations during the window of opportunity;
calculating a plurality of average velocities by averaging the plurality of raw velocities over a unit of time;
comparing the plurality of average velocities to a set of threshold values;
detecting that a first average speed has fallen below a first threshold, indicating the opening of a first window of opportunity;
detecting that a second average speed has fallen below a second threshold, indicating the opening of a second window of opportunity;
providing hints for the first and second windows of opportunity;
determining whether to perform the first subset of pending or impending background operations during the first window of opportunity; and
and determining whether to perform the second subset of pending or impending background operations during the second window of opportunity. The method of claim 1,
initiating a host command among the plurality of host commands;
posting a completion input for each of the plurality of host commands; and
after executing each host command, updating the average rate. The method of claim 1,
wherein the host sends a host command to the SSD to program the host-programmable weight. The method of claim 1,
the host interface layer of the SSD provides the hints indicating that the window of opportunity is available to a Flash Translation Layer (FTL) of the SSD. The method of claim 1,
The threshold value is determined based on one or more of an application running on the host, a type of the SSD, and characteristics of the plurality of host commands. The method of claim 1,
detecting that the average velocity exceeds the threshold value, indicating the closing of the opportunity window; and
The method further comprising the step of suspending or curtailing active background operations. flash media; and
a controller comprising a host interface layer, a flash translation layer, and a flash interface layer;
The host interface layer is:
Receive a plurality of host commands from the host to access the storage medium of the SSD;
monitor a raw speed to perform the plurality of host commands based on the amount of data transferred between the host and the SSD;
compute an average velocity by averaging the raw velocity over a unit of time using host-programmable weights;
compare the average speed to a threshold;
detect that the average velocity falls below the threshold, indicating an opening of a window of opportunity; and
and provide hints for the opportunity window to the flash transformation layer;
the flash transformation layer is configured to determine whether to perform pending or impending background operations during the opportunity window;
wherein the flash interface layer is configured to perform data transfer to or from the flash medium according to the plurality of host commands;
The host interface layer is:
calculate the plurality of average velocities by averaging the plurality of raw velocities over a unit of time;
compare the plurality of average velocities to a set of threshold values;
detect that a first average velocity, indicating opening of a first window of opportunity, falls below a first threshold;
detect that a second average velocity, indicating opening of a second window of opportunity, falls below a second threshold;
provide hints for the first and second opportunity windows;
determine whether to perform the first subset of pending or impending background operations during the first window of opportunity; and
and determine whether to perform the second subset of pending or impending background operations during the second window of opportunity. 8. The method of claim 7,
The host interface layer is:
initiating a host command of the plurality of host commands;
post a completion input for each of the plurality of host commands; and
SSD configured to update the average speed after executing each host command. 8. The method of claim 7,
wherein the host sends a host command to the SSD to program the host-programmable weight. 8. The method of claim 7,
the host interface layer is configured to detect that the average speed, indicating the closing of the opportunity window, exceeds the threshold value, and the flash translation layer is configured to suspend or reduce active background operations.

Images